About
Riccardo Kiefer is a mathematical engineer interested in applied modelling, scientific computing, data-driven analysis, and quantitative approaches to complex systems. This website brings together publications, academic projects, and selected documents in a clear personal archive.
I am a mathematical engineer particularly interested in the theoretical study of problems that can lead to applied solutions. My interests are broad, but they are connected by the study of complex systems, where multiple variables interact in nonlinear ways and give rise to emergent behaviours. I am especially drawn to models that extend beyond a purely mathematical setting and can help interpret phenomena in fields such as environmental science, economics, neuroscience, sociology and psychology.
This is the reason why my recent academic work focused on data-driven surrogate modelling for street-scale NOx concentration prediction, combining simulation-derived data, meteorological variables and statistical methods. This experience strengthened my background in computational modelling, statistical learning, optimization, scientific computing and high-performance computing.
I completed my MSc in Mathematical Engineering at Politecnico di Torino, with a research internship at École Centrale de Lyon. This path trained me to approach problems with analytical rigor and intellectual flexibility, move across different mathematical and computational frameworks, and formulate, prototype and critically test solutions across heterogeneous application domains.
MSc thesis project on data driven approaches for predicting street scale passive scalar concentrations using simulation derived inputs from the SIRANE model. The work focused on dataset construction, feature engineering and the comparison of statistical and machine learning models for concentration prediction at local urban scale.
Research activity focused on urban air pollution modelling, with attention to meteorological forcing, street geometry, emission scenarios and local scale concentration fields. The work connected simulation outputs with statistical analysis and data driven modelling strategies.
BSc thesis on the theoretical foundations of artificial neural networks and deep learning. The work introduced artificial neurons, perceptrons, sigmoid neurons, feedforward architectures, gradient descent, stochastic gradient descent and the backpropagation algorithm. It also discussed handwritten digit recognition, deep neural networks, the vanishing gradient problem and convolutional neural network architectures for image analysis.
Selected MSc projects developed during the Mathematical Engineering programme, covering optimization, stochastic systems, network dynamics, machine learning, logistics analytics, quantitative risk and distributed ledger architectures.
Numerical optimization project on equality constrained quadratic programming. The work derived the KKT optimality system and compared direct solvers, GMRES with and without preconditioning, Schur complement approaches and the null space method across increasingly large sparse problem instances. The analysis focused on accuracy, residuals, computational time, scalability and numerical robustness.
Stochastic optimization project on heuristic scenario reduction for two stage stochastic programming. The work implemented aggregation based and Wasserstein distance based reduction methods within a scenario tree framework and tested them on newsvendor and assembly to order models solved with Gurobi. The analysis compared in sample and out of sample stability, solution quality, reduced tree size and computational scalability.
Presentation and paper analysis on sourcing strategies under supply risk, based on an approximate dynamic programming approach. The work examined finite horizon procurement decisions with regular and emergency suppliers, stochastic supplier reliability, Markovian rating transitions, inventory dynamics and lead time effects. The presentation compared dynamic programming and approximate dynamic programming formulations, with attention to post decision states, exploration vs exploitation, stepsize rules and aggregation levels.
Business analytics project on the Capacitated Vehicle Routing Problem. The work implemented and compared two constructive heuristics: Sweep clustering with Nearest Neighbour routing and the Clarke Wright savings algorithm, followed by an intra route Tabu Search improvement phase. The methods were tested on both randomly generated instances and benchmark CVRP data, comparing route costs, capacity usage, computational time and distance from the known optimum.
A set of MSc coursework projects on dynamical processes over networks, combining graph theoretic models, flow optimization, congestion games, consensus dynamics, stochastic transitions and strategic interaction on graphs. The assignments covered max flow/min cut analysis, Wardrop equilibria, Braess paradox, toll design, convergence to consensus, bottleneck effects, Cheeger type bounds, key player identification and game theoretic dynamics on networks.
Full coursework documents available upon request.
Coursework projects on discrete time Markov chains, stationary distributions, periodicity, hitting and return times, non homogeneous Poisson processes, gambler’s ruin models and martingale techniques. The work included analytical derivations, exit distribution computations and Monte Carlo simulations, with a theoretical extension based on Doob decomposition, stopping times and optional stopping arguments for biased random walks.
Full coursework documents available upon request.
Python project on automatic face recognition using the eigenfaces method. The work implemented a PCA based pipeline including dataset preprocessing, train/test splitting, mean face computation, covariance analysis, dimensionality reduction, projection onto the eigenface subspace and nearest neighbour classification under different recognition thresholds.
Presentation on distributionally robust international portfolio optimization under asset return and exchange rate uncertainty. The work analysed a worst case mean CVaR framework, support set construction under no arbitrage constraints, ambiguity sets based on moment uncertainty and semi definite programming reformulations. Numerical experiments compared robust and benchmark strategies through rolling window backtesting over total, crisis and stable market periods.
Presentation on Corda as a permissioned distributed ledger platform for financial services. The work examined Corda’s transaction model, state objects, smart contracts, flows, vaults, notary services and double spending prevention, highlighting the differences between Corda, Bitcoin, Ethereum and broader distributed ledger technologies.
Ongoing patent related activity concerning the development and assessment of a biocomposite material for agricultural fertilizing applications. The work involves technical documentation, patent landscaping, preliminary market and competitor analysis, and the evaluation of potential application scenarios within the agricultural and biostimulant sectors.
Further technical details are not publicly disclosed due to confidentiality constraints.
Design and maintenance of a Raspberry Pi based homelab using Docker, Caddy, Cloudflare, VPN based private access, service isolation and automated backups. The project explores reverse proxying, public/private service separation, federated services and lightweight infrastructure management.
Participation in BitPolito activities focused on cryptography, Bitcoin, blockchain systems and decentralized technologies. The experience contributed to the development of technical knowledge in applied cryptography, distributed systems, protocol level thinking and educational dissemination within a university based technical community.
Selected reports and presentations are linked directly from the project cards. Some coursework documents are intentionally not published in full and are available upon request.
For academic or professional inquiries, please contact me by email or through LinkedIn.
Email: [email protected]
LinkedIn: Riccardo Kiefer